fluid states: fix massFraction() method for the case where the average molar mass is negative

yeah, that's really exotic, but this occationally happens to me. (if i
force all mole fractions to be positive, I get worse convergence
rates.)

git-svn-id: svn://svn.iws.uni-stuttgart.de/DUMUX/dumux/trunk@7443 2fb0f335-1f38-0410-981e-8018bf24f1b0

dumux/material/fluidstates/compositionalfluidstate.hh

@@ -87,10 +87,10 @@ public:
     Scalar massFraction(int phaseIdx, int compIdx) const
     {
         return
-            sumMoleFractions_[phaseIdx]
+            std::abs(sumMoleFractions_[phaseIdx])
             * moleFraction_[phaseIdx][compIdx]
             * FluidSystem::molarMass(compIdx)
-            / std::max(1e-10, averageMolarMass_[phaseIdx]);
+            / std::max(1e-40, std::abs(averageMolarMass_[phaseIdx]));
     }
 
     /*!

dumux/material/fluidstates/nonequilibriumfluidstate.hh

@@ -86,10 +86,10 @@ public:
     Scalar massFraction(int phaseIdx, int compIdx) const
     {
         return
-            sumMoleFractions_[phaseIdx]
+            std::abs(sumMoleFractions_[phaseIdx])
             * moleFraction_[phaseIdx][compIdx]
             * FluidSystem::molarMass(compIdx)
-            / std::max(1e-10, averageMolarMass_[phaseIdx]);
+            / std::max(1e-40, std::abs(averageMolarMass_[phaseIdx]));
     }
 
     /*!